Heat: high-efficiency simulation for thermal ablation therapy

Percutaneous thermal ablation is increasingly popular but still suffers from a complex preoperative planning, especially with multiple needles. Existing planning methods either use theoretical ablation shapes for faster estimates or are computationally intensive when incorporating realistic thermal propagation. This paper introduces a multi-resolution approach that accelerates thermal propagation simulation, enabling users to adjust ablation parameters and see the results in interactive time.

For static needle positions, a high-resolution simulation based on GPU-accelerated implementation of the Pennes bioheat equation is used. During user interaction, intermediate frames display a lower-resolution estimation of the ablated volume. Two methods are compared, based on GPU-accelerated reimplementations of finite difference and lattice Boltzmann approaches. A parameter study was conducted to identify the optimal balance between speed and accuracy for the low- and high-resolution frames. The chosen parameters are finally tested in multi-needle scenarios to validate the interactive capability in this context.

Tested with percutaneous radiofrequency data, our multi-resolution method significantly reduces computation time while maintaining good accuracy compared to the reference simulation. For high-resolution frames, we can reach up to 5.8 fps, while for intermediate low-resolution frames we can reach a frame rate of 32 fps with less than 20% loss of accuracy.

This multi-resolution approach allows for smooth interaction with multiple needles, with instant visualization of the predicted ablation volume, in the context of percutaneous radiofrequency treatments. It could also be applied to automated planning, reducing the time required for iterative adjustments.